CN112100333A - Online recommendation system based on deep learning and knowledge graph fusion - Google Patents
Online recommendation system based on deep learning and knowledge graph fusion Download PDFInfo
- Publication number
- CN112100333A CN112100333A CN202010979963.6A CN202010979963A CN112100333A CN 112100333 A CN112100333 A CN 112100333A CN 202010979963 A CN202010979963 A CN 202010979963A CN 112100333 A CN112100333 A CN 112100333A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- user
- product
- entity
- recommendation system
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 67
- 230000004927 fusion Effects 0.000 title claims abstract description 27
- 238000013136 deep learning model Methods 0.000 claims abstract description 29
- 230000005540 biological transmission Effects 0.000 claims abstract description 8
- 238000000034 method Methods 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 10
- 230000007246 mechanism Effects 0.000 claims description 4
- 230000007115 recruitment Effects 0.000 claims description 3
- 238000000926 separation method Methods 0.000 claims description 3
- 230000003993 interaction Effects 0.000 abstract description 3
- 230000006399 behavior Effects 0.000 description 11
- 238000001914 filtration Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 238000007418 data mining Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000035800 maturation Effects 0.000 description 2
- 230000006641 stabilisation Effects 0.000 description 2
- 238000011105 stabilization Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Computational Linguistics (AREA)
- Finance (AREA)
- Artificial Intelligence (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to an online recommendation system based on deep learning and knowledge graph fusion, which comprises: the knowledge graph building module is configured to build a knowledge graph according to the user entity, the product entity, the channel entity and the incidence relation among the entities; a model building module configured to build a deep learning model from the knowledge graph, the deep learning model comprising a deep learning unit and a linear sub-model; the recommending module is configured to predict by adopting the deep learning model according to the user entity, the product entity and/or the channel entity to obtain a recommended predicted value and complete online recommendation according to the recommended predicted value; and the data transmission module and the client are configured to complete transmission and human-computer interaction of all data in the online recommendation system. The online recommendation system provided by the invention realizes the supply of products with diversity and relevance to users, and solves the problems of insufficient freshness and single freshness of recommended products caused by data sparseness and cold start.
Description
Technical Field
The invention relates to the technical field of data mining and online recommendation, in particular to an online recommendation system based on deep learning and knowledge map fusion.
Background
With the rapid development of the internet, various information is growing explosively, so that the occurrence of 'information overload' is caused, and the recommendation system is generated against the problem. The recommendation system can provide good decision support and personalized service for users. At present, a collaborative filtering algorithm recommends a product of interest for a user according to interest or preference by methods such as data mining and machine learning.
In the prior art, most recommendation systems recommend products to users by using expert rules or collaborative filtering algorithms, but the expert rules often cannot fully consider the relationship between the users and the recommended products, and the collaborative filtering algorithms can understand the intentions of the users and recommend favorite products by using historical information of the users, so that the problems of data sparseness, cold start and the like are often encountered although incomplete or inaccurate content analysis is avoided to a certain extent, and the problems of excessive similarity of recommended products and insufficient diversity of recommended products are encountered in both collaborative filtering based on users and collaborative filtering based on product contents.
Therefore, an online recommendation system based on deep learning and knowledge graph fusion is needed to provide products with diversity and relevance for users, so as to solve the problems of insufficient freshness and over-singleness of recommended products caused by data sparseness and cold start in the prior art.
Disclosure of Invention
The invention aims to provide an online recommendation system based on deep learning and knowledge graph fusion, which provides products with diversity and relevance for users, and solves the problems of insufficient freshness and single freshness of recommended products caused by data sparseness and cold start in the prior art.
In order to solve the problems in the prior art, the invention provides an online recommendation system based on deep learning and knowledge graph fusion, which comprises:
the knowledge graph building module is configured to build a knowledge graph according to the user entity, the product entity, the channel entity and the incidence relation among the entities;
a model building module configured to build a deep learning model from the knowledge graph, the deep learning model comprising a deep learning unit and a linear sub-model;
the recommending module is configured to predict by adopting the deep learning model according to the user entity, the product entity and/or the channel entity to obtain a recommended predicted value, and complete online recommendation according to the recommended predicted value;
the data transmission module is configured to complete transmission of all data in the online recommendation system; and
and the client is configured to enable the user to interact with the online recommendation system.
Optionally, in the deep learning and knowledge graph fusion-based online recommendation system, the association relationship between the entities includes:
the method comprises the steps of obtaining an association relationship between a user entity and a product entity, an association relationship between the user entity and a channel entity, an association relationship between the product entity and the channel entity, an association relationship between different user entities, an association relationship between different product entities and an association relationship between different channel entities.
Optionally, in the online recommendation system based on deep learning and knowledge graph fusion,
the user entity comprises basic information of the user and user tag data, wherein the user tag data comprises the period of the user, user preference and social network;
the product entity comprises basic information of a product and product label data, wherein the product label data comprises a product risk grade, a product recruitment mode and a product fund company;
the channel entities include channel entities for online transactions and offline transactions.
Optionally, in the online recommendation system based on deep learning and knowledge graph fusion, the method for establishing the deep learning model is as follows:
respectively extracting continuous features and class-type features in the knowledge graph;
combining the extracted features;
performing abstract processing and linear division on the combined features to obtain different types of data;
deep learning is carried out on the obtained data of different types to form a deep learning unit;
and obtaining a deep learning model from the results of the deep learning unit and the linear submodel.
Optionally, in the online recommendation system based on deep learning and knowledge graph fusion,
the class-type features include high-dimensional sparse features and low-dimensional dense features.
Optionally, in the online recommendation system based on deep learning and knowledge graph fusion, the deep learning unit converts the extracted high-dimensional sparse features into low-dimensional dense features to increase an association relationship among a user entity, a product entity and a channel entity.
Optionally, in the online recommendation system based on deep learning and knowledge graph fusion, the way of establishing the linear submodel is as follows:
respectively extracting continuous features and low-dimensional dense features in the knowledge graph;
performing box separation on the extracted continuous features;
and establishing a linear sub-model according to the continuous features subjected to binning processing and the extracted low-dimensional dense features.
Optionally, in the online recommendation system based on deep learning and knowledge graph fusion, before obtaining the deep learning model, the method further includes the following steps: and optimizing and adjusting the parameters of the deep learning unit and the linear submodel by adopting a loss function.
Optionally, in the online recommendation system based on deep learning and knowledge graph fusion, the method for optimizing and adjusting the parameters of the deep learning unit and the linear submodel is as follows:
weighting the logarithms of the deep learning unit and the linear submodel to obtain a model predicted value;
and pushing the model prediction value into the loss function, dynamically adjusting parameters of a deep learning unit and a linear submodel through the minimum batch random gradient of the loss function, and determining appropriate parameters to form a final deep learning model.
Optionally, in the online recommendation system based on deep learning and knowledge graph fusion, the client is further configured to provide a recommendation result of online recommendation for the user when the user triggers the field mechanism.
Compared with the prior art, the invention has the following advantages:
1. by fusing the user entity, the product entity, the channel entity and the incidence relation among the entities in the knowledge graph in the deep learning process, the practicability of the data is greatly improved, and the problem of data sparsity is effectively relieved;
2. the problems of cold start of users and cold start of products are effectively solved;
3. by using a deep learning technology, the attributes of the product and the behavior information of the user on the product are fully utilized, the characteristics which often appear simultaneously are learned by using a linear sub-model, the accurate recommendation of the product is realized, and new characteristic combinations which hardly appear before the deep learning unit is used for learning, the diversified recommendation of the product is realized, so that the product recommended by the invention can accurately accord with the interest preference of the user, and meanwhile, the freshness of the user can be maintained by recommending some new product combinations;
4. the interpretability of the recommendation result is enhanced by connecting the historical records of the user with the recommendation result;
5. the user can more simply and conveniently acquire the information of interested products on the client, and the use experience of the user is greatly improved.
Drawings
FIG. 1 is a block diagram of an online recommendation system provided by an embodiment of the present invention;
fig. 2 is a flowchart of establishing a deep learning model according to an embodiment of the present invention.
Detailed Description
The following describes in more detail embodiments of the present invention with reference to the schematic drawings. The advantages and features of the present invention will become more apparent from the following description. It is to be noted that the drawings are in a very simplified form and are not to precise scale, which is merely for the purpose of facilitating and distinctly claiming the embodiments of the present invention.
Hereinafter, if the method described herein comprises a series of steps, the order of such steps presented herein is not necessarily the only order in which such steps may be performed, and some of the described steps may be omitted and/or some other steps not described herein may be added to the method.
In an actual application scene, interaction information of a user and a product only exists if the user has past behavior on the product, and for a rich product list, the user often only has interaction information on a small part of products, so that data can be quite sparse. The use of such a small amount of observed data to predict a large amount of unknown information greatly increases the risk of model overfitting, and the online operation effect of the system is often poor.
When a new user enters the existing recommendation system, the system only contains basic attribute information of the user, and the behavior data of the user is lacked, so that the future preference of the user cannot be predicted according to the historical behavior of the user. When a new product enters the product library, the data of the behavior related to the new product is lacked, and the new product cannot be recommended to users who may be interested in the new product, so that the problems of cold start of the users and cold start of the products can be caused.
Therefore, it is necessary to provide an online recommendation system based on deep learning and knowledge graph fusion, as shown in fig. 1 and 2, fig. 1 is a block diagram of the online recommendation system provided by the embodiment of the present invention; fig. 2 is a flowchart of establishing a deep learning model according to an embodiment of the present invention. The online recommendation system comprises:
the knowledge graph building module is configured to build a knowledge graph according to the user entity, the product entity, the channel entity and the incidence relation among the entities;
a model building module configured to build a deep learning model from the knowledge graph, the deep learning model comprising a deep learning unit and a linear sub-model;
the recommending module is configured to predict by adopting the deep learning model according to the user entity, the product entity and/or the channel entity to obtain a recommended predicted value, and complete online recommendation according to the recommended predicted value;
the data transmission module is configured to complete transmission of all data in the online recommendation system; and
and the client is configured to enable the user to interact with the online recommendation system.
According to the method, the user entity, the product entity, the channel entity and the incidence relation among the entities in the knowledge graph are fused in the deep learning process, so that the usability of data is greatly improved, and the problem of data sparsity is effectively solved; the problems of cold start of users and cold start of products are effectively solved; by using a deep learning technology, the attributes of the product and the behavior information of the user on the product are fully utilized, the characteristics which often appear simultaneously are learned by using a linear sub-model, the accurate recommendation of the product is realized, and new characteristic combinations which hardly appear before the deep learning unit is used for learning, the diversified recommendation of the product is realized, so that the product recommended by the invention can accurately accord with the interest preference of the user, and meanwhile, the freshness of the user can be maintained by recommending some new product combinations; the interpretability of the recommendation result is enhanced by connecting the historical records of the user with the recommendation result; moreover, the user can acquire the information of the interested product more simply and conveniently on the client, and the use experience of the user is greatly improved.
Specifically, the dynamic relationship characteristics among the entities are combined with the basic information of the user and the structured data information such as the browsing and purchasing records of the user history, the user intention is analyzed, and the potential needs and the intention of the user are mined. The established knowledge graph can be used for mining the rich relations such as the association relation among users, the product preference of the users, the similarity among products, the channel preference of the users and the like, and the relation data is used as the original characteristics and introduced into a deep learning model, so that the problem of cold start of the users and the products can be effectively solved. When accurate recommendation is carried out, if a new user enters an online channel, characteristic data such as historical operation behaviors of the user are seriously lost, the recommendation result is often seriously distorted only depending on personal basic information of the user when model analysis is carried out, the knowledge graph introduces relationship data among user entities, old users similar to the new user can be obtained, and products more conforming to the preference of the old users are obtained according to the basic information of the new user and the historical information of the old users similar to the new user. Similarly, a new product only has attribute data of the product, and the historical browsing and purchasing data of the product is completely absent, when the product is recommended to the user, the recommendation priority of the newly entered product is extremely low, so that the product cannot be recommended to the user, the data accumulation of the product is slow, and the product needs to have enough data to be effectively recommended, so that the product needs to be accumulated for a longer time. The relationship data among the product entities in the knowledge graph can also relate the new product and the old product, and the effectiveness of new product recommendation is improved according to the attribute data of the new product and the data such as the user transaction behavior of the old product similar to the new product.
To sum up, the association relationship among the entities is obtained according to the user entity, the product entity and the channel entity when the knowledge graph is constructed, and the association relationship among the entities includes, for example: the method comprises the steps of obtaining an association relationship between a user entity and a product entity, an association relationship between the user entity and a channel entity, an association relationship between the product entity and the channel entity, an association relationship between different user entities, an association relationship between different product entities and an association relationship between different channel entities. In one embodiment, the most common association relationship between different user entities may be a relationship of the same kind of relatives, friends and consumption habits, the most common association relationship between different product entities may be a relationship of the hot-market products and the same kind of products, and the most common association relationship between different channel entities may be a relationship of the online channel and the offline channel.
Further, the content contained in the user entity, the product entity and the channel entity is specifically as follows: the user entity comprises basic information of the user and user tag data, wherein the user tag data comprises the period of the user, user preference and social network; the period comprises an introduction period, a maturation period, a stabilization period and a loss period, and the users are divided into the introduction period, the maturation period, the stabilization period, the loss period and the like according to the conditions of online operation behaviors of the users, transaction trends of offline products and the like; the user preference is obtained from historical purchase records and/or historical browsing records and the like of the user, and the social network is a network formed by public and/private social relations of the user. The invention can also establish user label data according to the channel preference of the user so as to enrich the portrait of the user and improve the relationship between user entities. The product entity comprises basic information of a product and product label data, wherein the product label data comprises a product risk grade, a product recruitment mode and a product fund company; the channel entities comprise channel entities of online transactions and offline transactions, the online transactions comprise different transaction APPs, transaction webpages and the like, the offline transactions comprise website transactions, counter transactions and the like, and various user entities and product entities can have unique channel preferences.
In an actual situation, because the interests and hobbies of the user are limited, the interactive information between the user and the product is less, the data is in a high-dimensional sparse state, the operation efficiency of the existing recommendation model can be reduced if the data are directly used, the effect of the existing recommendation model is not accurate due to the fact that the information contained in the data is less, the deep learning unit can convert the high-dimensional sparse features into the low-dimensional dense features through Embedding, and the features can be combined in a variety of ways in the conversion process. The deep learning unit can introduce all features into fitting of a model function, and the process can blur some direct causal relationships, generalize and generate some indirect and possible correlations, and can realize diversified recommendation of products. The deep learning model combines the results of the linear submodel part and the deep learning unit to obtain the final recommendation result for the user, so that the final result has the characteristics of accuracy and diversity.
Specifically, referring to fig. 2, the deep learning model realizes accurate recommendation of a product by using massive data containing rich information, and the deep learning model is established in the following manner:
respectively extracting continuous features and class-type features in the knowledge graph;
combining the extracted features;
performing abstract processing and linear division on the combined features to obtain different types of data;
deep learning is carried out on the obtained data of different types to form a deep learning unit;
and obtaining a deep learning model from the results of the deep learning unit and the linear submodel.
Typically, the class-type features include high-dimensional sparse features and low-dimensional dense features. And the deep learning unit converts the extracted high-dimensional sparse features into low-dimensional dense features so as to increase the incidence relation among the user entity, the product entity and the channel entity.
Further, the way of establishing the linear submodel is as follows:
respectively extracting continuous features and low-dimensional dense features in the knowledge graph;
performing box separation on the extracted continuous features;
and establishing a linear sub-model according to the continuous features subjected to binning processing and the extracted low-dimensional dense features.
Specifically, the linear submodel mainly uses low-dimensional dense features and continuous data discretized data. The low-dimensional dense features belong to the class type features, the class type features comprise data such as user risk levels, asset levels and education degrees, and the continuous data discretization scientifically boxes data such as ages, incomes, assets and product transactions according to modeling targets, so that the operation speed of the linear sub-model can be increased, and the stability of the model can be improved. The linear submodel finds the correlation between the user and the characteristics from the historical data, and the generated recommendation generally recommends products directly related to the products with which the user has interacted, so that the recommendation accuracy is improved.
Preferably, before obtaining the deep learning model, the method further comprises the following steps: and optimizing and adjusting the parameters of the deep learning unit and the linear submodel by adopting a loss function.
Further, the method for optimizing and adjusting the parameters of the deep learning unit and the linear submodel is as follows:
weighting the logarithms of the deep learning unit and the linear submodel to obtain a model predicted value;
and pushing the model prediction value into the loss function, dynamically adjusting parameters of a deep learning unit and a linear submodel through the minimum batch random gradient of the loss function, and determining appropriate parameters to form a final deep learning model.
Preferably, the client is further configured to provide a recommendation result of online recommendation to the user when the user triggers the field mechanism. Specifically, after the user logs in at the client, the identification information of the user can be generated, when the operation behavior of the user triggers the column mechanism, the recommendation result can be displayed on the corresponding page of the client, the recommendation result contains a product or some functional services, so that the user can acquire the information of the interested product more simply and conveniently on the client, the use experience of the user is greatly improved, and the exposure of the product is also improved.
In summary, compared with the prior art, the invention has the following advantages:
1. by fusing the user entity, the product entity, the channel entity and the incidence relation among the entities in the knowledge graph in the deep learning process, the practicability of the data is greatly improved, and the problem of data sparsity is effectively relieved;
2. the problems of cold start of users and cold start of products are effectively solved;
3. by using a deep learning technology, the attributes of the product and the behavior information of the user on the product are fully utilized, the characteristics which often appear simultaneously are learned by using a linear sub-model, the accurate recommendation of the product is realized, and new characteristic combinations which hardly appear before the deep learning unit is used for learning, the diversified recommendation of the product is realized, so that the product recommended by the invention can accurately accord with the interest preference of the user, and meanwhile, the freshness of the user can be maintained by recommending some new product combinations;
4. the interpretability of the recommendation result is enhanced by connecting the historical records of the user with the recommendation result;
5. the user can more simply and conveniently acquire the information of interested products on the client, and the use experience of the user is greatly improved.
The above description is only a preferred embodiment of the present invention, and does not limit the present invention in any way. It will be understood by those skilled in the art that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. An online recommendation system based on deep learning and knowledge graph fusion is characterized by comprising:
the knowledge graph building module is configured to build a knowledge graph according to the user entity, the product entity, the channel entity and the incidence relation among the entities;
a model building module configured to build a deep learning model from the knowledge graph, the deep learning model comprising a deep learning unit and a linear sub-model;
the recommending module is configured to predict by adopting the deep learning model according to the user entity, the product entity and/or the channel entity to obtain a recommended predicted value, and complete online recommendation according to the recommended predicted value;
the data transmission module is configured to complete transmission of all data in the online recommendation system; and
and the client is configured to enable the user to interact with the online recommendation system.
2. The deep learning and knowledge-graph fusion based online recommendation system of claim 1, wherein the association between entities comprises:
the method comprises the steps of obtaining an association relationship between a user entity and a product entity, an association relationship between the user entity and a channel entity, an association relationship between the product entity and the channel entity, an association relationship between different user entities, an association relationship between different product entities and an association relationship between different channel entities.
3. The deep learning and knowledge-graph fusion based online recommendation system of claim 1,
the user entity comprises basic information of the user and user tag data, wherein the user tag data comprises the period of the user, user preference and social network;
the product entity comprises basic information of a product and product label data, wherein the product label data comprises a product risk grade, a product recruitment mode and a product fund company;
the channel entities include channel entities for online transactions and offline transactions.
4. The deep learning and knowledge-graph fusion based online recommendation system of claim 1, wherein the deep learning model is built by:
respectively extracting continuous features and class-type features in the knowledge graph;
combining the extracted features;
performing abstract processing and linear division on the combined features to obtain different types of data;
deep learning is carried out on the obtained data of different types to form a deep learning unit;
and obtaining a deep learning model from the results of the deep learning unit and the linear submodel.
5. The deep learning and knowledge-graph fusion based online recommendation system of claim 4,
the class-type features include high-dimensional sparse features and low-dimensional dense features.
6. The deep learning and knowledge-graph fusion based online recommendation system of claim 5, wherein the deep learning unit converts the extracted high-dimensional sparse features into low-dimensional dense features to increase the incidence relation between user entities, product entities and channel entities.
7. The deep learning and knowledge-graph fusion based online recommendation system of claim 5, wherein the linear submodel is established by:
respectively extracting continuous features and low-dimensional dense features in the knowledge graph;
performing box separation on the extracted continuous features;
and establishing a linear sub-model according to the continuous features subjected to binning processing and the extracted low-dimensional dense features.
8. The deep learning and knowledge-graph fusion based online recommendation system of claim 4, further comprising, before obtaining the deep learning model, the steps of: and optimizing and adjusting the parameters of the deep learning unit and the linear submodel by adopting a loss function.
9. The deep learning and knowledge-graph fusion based online recommendation system of claim 8, wherein the parameters of the deep learning unit and the linear submodel are optimally adjusted by:
weighting the logarithms of the deep learning unit and the linear submodel to obtain a model predicted value;
and pushing the model prediction value into the loss function, dynamically adjusting parameters of a deep learning unit and a linear submodel through the minimum batch random gradient of the loss function, and determining appropriate parameters to form a final deep learning model.
10. The deep learning and knowledge-graph fusion based online recommendation system of claim 1, wherein the client is further configured to provide a recommendation result of online recommendation to the user when the user triggers a field mechanism.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010979963.6A CN112100333A (en) | 2020-09-17 | 2020-09-17 | Online recommendation system based on deep learning and knowledge graph fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010979963.6A CN112100333A (en) | 2020-09-17 | 2020-09-17 | Online recommendation system based on deep learning and knowledge graph fusion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112100333A true CN112100333A (en) | 2020-12-18 |
Family
ID=73759906
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010979963.6A Pending CN112100333A (en) | 2020-09-17 | 2020-09-17 | Online recommendation system based on deep learning and knowledge graph fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112100333A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113158052A (en) * | 2021-04-23 | 2021-07-23 | 平安银行股份有限公司 | Chat content recommendation method and device, computer equipment and storage medium |
| CN116720786A (en) * | 2023-08-01 | 2023-09-08 | 中国科学院工程热物理研究所 | KG and PLM fusion assembly quality stability prediction method, system and medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107194011A (en) * | 2017-06-23 | 2017-09-22 | 重庆邮电大学 | A kind of position prediction system and method based on social networks |
| US20180089593A1 (en) * | 2016-09-26 | 2018-03-29 | Acusense Technologies, Inc. | Method and system for an end-to-end artificial intelligence workflow |
| US20180150522A1 (en) * | 2016-08-09 | 2018-05-31 | Ripcord, Inc. | Systems and methods for contextual retrieval and contextual display of records |
| CN108875827A (en) * | 2018-06-15 | 2018-11-23 | 广州深域信息科技有限公司 | A kind of method and system of fine granularity image classification |
| CN109684548A (en) * | 2018-11-30 | 2019-04-26 | 内江亿橙网络科技有限公司 | A kind of data recommendation method based on user's map |
| CN110647620A (en) * | 2019-09-23 | 2020-01-03 | 中国农业大学 | Knowledge graph representation learning method based on confidence hyperplane and dictionary information |
| CN110795571A (en) * | 2019-10-24 | 2020-02-14 | 南宁师范大学 | Cultural tourism resource recommendation method based on deep learning and knowledge graph |
| CN110866190A (en) * | 2019-11-18 | 2020-03-06 | 支付宝(杭州)信息技术有限公司 | Method and device for training neural network model for representing knowledge graph |
| CN111190968A (en) * | 2019-12-16 | 2020-05-22 | 北京航天智造科技发展有限公司 | Data preprocessing and content recommendation method based on knowledge graph |
| CN111191851A (en) * | 2020-01-03 | 2020-05-22 | 中国科学院信息工程研究所 | Data center energy efficiency optimization method based on knowledge graph |
-
2020
- 2020-09-17 CN CN202010979963.6A patent/CN112100333A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180150522A1 (en) * | 2016-08-09 | 2018-05-31 | Ripcord, Inc. | Systems and methods for contextual retrieval and contextual display of records |
| US20180089593A1 (en) * | 2016-09-26 | 2018-03-29 | Acusense Technologies, Inc. | Method and system for an end-to-end artificial intelligence workflow |
| CN107194011A (en) * | 2017-06-23 | 2017-09-22 | 重庆邮电大学 | A kind of position prediction system and method based on social networks |
| CN108875827A (en) * | 2018-06-15 | 2018-11-23 | 广州深域信息科技有限公司 | A kind of method and system of fine granularity image classification |
| CN109684548A (en) * | 2018-11-30 | 2019-04-26 | 内江亿橙网络科技有限公司 | A kind of data recommendation method based on user's map |
| CN110647620A (en) * | 2019-09-23 | 2020-01-03 | 中国农业大学 | Knowledge graph representation learning method based on confidence hyperplane and dictionary information |
| CN110795571A (en) * | 2019-10-24 | 2020-02-14 | 南宁师范大学 | Cultural tourism resource recommendation method based on deep learning and knowledge graph |
| CN110866190A (en) * | 2019-11-18 | 2020-03-06 | 支付宝(杭州)信息技术有限公司 | Method and device for training neural network model for representing knowledge graph |
| CN111190968A (en) * | 2019-12-16 | 2020-05-22 | 北京航天智造科技发展有限公司 | Data preprocessing and content recommendation method based on knowledge graph |
| CN111191851A (en) * | 2020-01-03 | 2020-05-22 | 中国科学院信息工程研究所 | Data center energy efficiency optimization method based on knowledge graph |
Non-Patent Citations (3)
| Title |
|---|
| DACREMA, M.F等: "Are We Really Making Much Progress? A Worrying Analysis of Recent Neural Recommendation Approaches", 《ASSOCIATION FOR COMPUTING MACHINERY》, 16 July 2019 (2019-07-16), pages 101 - 109 * |
| 汪加林: "基于用户偏好的深度学习推荐系统", 《中国优秀硕士学位论文全文数据库·信息科技辑》, no. 06, 15 June 2019 (2019-06-15), pages 1 - 74 * |
| 黄立威等: "基于深度学习的推荐系统研究综述", 《计算机学报》, vol. 41, no. 07, 5 March 2018 (2018-03-05), pages 1619 - 1647 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113158052A (en) * | 2021-04-23 | 2021-07-23 | 平安银行股份有限公司 | Chat content recommendation method and device, computer equipment and storage medium |
| CN116720786A (en) * | 2023-08-01 | 2023-09-08 | 中国科学院工程热物理研究所 | KG and PLM fusion assembly quality stability prediction method, system and medium |
| CN116720786B (en) * | 2023-08-01 | 2023-10-03 | 中国科学院工程热物理研究所 | KG and PLM fusion assembly quality stability prediction method, system and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kim et al. | Collaborative error-reflected models for cold-start recommender systems | |
| Ead et al. | Intelligent Systems of Machine Learning Approaches for developing E-services portals | |
| Yao et al. | Synergies between association rules and collaborative filtering in recommender system: An application to auto industry | |
| Cunha et al. | Selecting collaborative filtering algorithms using metalearning | |
| CN103886486A (en) | Electronic commerce recommending method based on support vector machine (SVM) | |
| De Maio et al. | Social media marketing through time‐aware collaborative filtering | |
| Zhu et al. | SI-News: Integrating social information for news recommendation with attention-based graph convolutional network | |
| Ambulgekar et al. | A survey on collaborative filtering: tasks, approaches and applications | |
| CN112100333A (en) | Online recommendation system based on deep learning and knowledge graph fusion | |
| Stephen et al. | Measures of similarity in memory-based collaborative filtering recommender system: A comparison | |
| CN112380434A (en) | Interpretable recommendation system method fusing heterogeneous information network | |
| Xia et al. | Exploiting temporal dynamics in product reviews for dynamic sentiment prediction at the aspect level | |
| Li | Accurate digital marketing communication based on intelligent data analysis | |
| Shi et al. | Extracting user influence from ratings and trust for rating prediction in recommendations | |
| Wangwatcharakul et al. | Dynamic collaborative filtering based on user preference drift and topic evolution | |
| Hong | Decrease and conquer-based parallel tensor factorization for diversity and real-time of multi-criteria recommendation | |
| Valdiviezo-Díaz et al. | A hybrid approach of recommendation via extended matrix based on collaborative filtering with demographics information | |
| Domingues et al. | Using contextual information from topic hierarchies to improve context-aware recommender systems | |
| Wang et al. | Consumer consumption behavioral model for business intelligence using artificial intelligence | |
| Gao et al. | Construction of digital marketing recommendation model based on random forest algorithm | |
| Gopalachari | DBT recommender: improved trustworthiness of ratings through de-biasing tendency of users | |
| Sreepada et al. | Revisiting tendency based collaborative filtering for personalized recommendations | |
| Padti et al. | Hybrid Movie Recommender System | |
| Lin et al. | MF-DMPC: Matrix factorization with dual multiclass preference context for rating prediction | |
| Malik et al. | Analysis of current recommendation techniques and evaluation metrics to design an improved book recommendation system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |